Machine Learning Models for Anxiety Detection and Prediction Using Perceived Control Data

Anxiety, as defined by the World Health Organization (WHO), is the intense and excessive feeling of fear and worry. It is considered one of the precursors of depression and other mental health conditions. Perceived control refers to the belief or perception that one has the ability to achieve positive outcomes through their own actions, and it is closely associated with mental health. Individuals with high levels of perceived control are strongly linked to good mental well-being and psychological health. We utilized an Android app that allowed users to estimate their level of control over a 'boing' sound after multiple interactions with the app. This data and other user behaviour data are extracted and used to generate Machine Learning models to predict symptoms of anxiety. We analyzed 401 samples, with 115 showing symptoms of anxiety and 286 not showing any symptoms. The models achieved up to 88% Mean ROC/AUC and a mean of 79.5% for Area Under the Precision-Recall curve with a 6-fold cross validation technique using the Random Forest algorithm. The results suggest a link between perception of control and anxiety, offering insights for further exploration.